Back in 2019, my previous gaming PC's motherboard fried due to a power surge. It was an old machine, but I liked it. It had an Intel i7 CPU, GTX 1070 and 16gb RAM. At the time, it was decent and handled many games perfectly fine.
Thankfully I had insurance for my machine, so I could get a new PC. The problem was, the hardware was irreplaceable, due to the chipset's age, and most new components of similar price and spec were incompatible. It wasn't really a problem, because it meant I could get an upgraded machine with modern components!
My current (then new) PC sports an AMD Ryzen 7 3800 processor, a Gigabyte RTX 2070 Super GPU (8bg) and 32gb RAM (it was 16 when I got it new, I added 16 more). Beefy, if I have to say so myself. I could run games like Cyberpunk 2077 at maxed-out graphics settings and the GPU temperatures were fine.
Some time, about a year or more ago - I cannot remember the exact time - my GPU fans started spinning up to maximum speed during gameplay, of things like Guild Wars 2 and other newer games. It didn't bother me too much at first, but over time it started to break my immersion and the loud fan spinning sounds began to bother me. I kept on installing new Nvidia drivers as they were released, but that didn't change anything. I monitored the GPU temperatures and it never really went above 85°C, which is not necessarily a problem. The thing that bothers me is the fan noise. It would ramp up to maximum fan speed for a few seconds, drop down to about 60-70%, then again ramp up to 100% and that constant ramping up is irritating.
You might say I could just use headphones, but sometimes I really don't want to.
The GPU idled around 40°C, which seems hot, but apparently that's normal. Even on a day like today, where it's 32°C outside, it currently idles at 36°C.
I have added some more case intake and exhaust fans, but that doesn't make a difference, only about 1, 2°C at most. I thought it was perhaps the poor airflow of my case, but it wasn't that at all. I added one additional intake fan at the front of the case and leave the main door open for better air flow, I took off the top panel and installed two more exhaust fans at the top. I noticed I accidentally installed the one back exhaust fan incorrectly during a cleaning session, so it was actually sucking air into the case. I fixed it and it didn't change much.
The other day I checked the GPU and it wasn't all that dirty, not a lot of dust build-up. I regularly clean my case, and I am wondering whether I should take the GPU apart and repaste it. I'm almost convinced that the thermal paste has gone hard and needs a replacement.
I've run a bunch of benchmarks in Cyberpunk 2077 today, with lowest possible graphics settings, VSync on and off, as well as DLSS (which is pretty cool!). I do not have any temperature improvements with highest settings and DLSS on ultra performance, but when I lower some settings, it stays under 80°C. However, the game looks like ass and I do not want to play like that.
I've first played it without updating Nvidia drivers, and then the exact same settings after I updated. I do not use Geforce Experience, because it is a piece of trash.
After some consideration, I decided to open the card and take a peek at the state of the thermal pads and paste. I've never done this before, so I spent some time looking online for information on repasting a graphics card. I purchased the necessary components (new thermal paste and thermal pads) and got cracking. At first I only purchased 1mm thick thermal pads, and applied that on the VRAM pads I accidentally tore when I opened the card. I measured incorrectly and they prevent the graphics card's chip from making contact with the heatsink, so my idle temperatures went up to 77°C, which made me shut the machine down immediately!
After measuring the VRAM pad thickness again, I found them to be 0,5mm thick, so I had to order another packet of thermal pads online. My usual computer store was closed by the time I was doing this, so I had to wait for two days for delivery. It's not so bad to wait for two days, because we are currently having about 10 hours of power cuts a day, so it's not like I can do anything that requires electricity anyway...
After reapplying the thermal paste and the 0,5mm pads on the VRAM modules, I put the card back together carefully and inserted it into my PC. I admit I was somewhat nervous, because maybe I fucked something up, but the machine started up fine and I fired up MSI Afterburner to monitor the temperatures.
A whopping 30°C idle!
What magic, what relief!
I tinkered with the fan curve a bit in MSI Afterburner to make sure it stays nice and cool during load and fired up some benchmarking software to test with. The same software that was making my card max out before the new paste and pads was now running a decent 60°C.
Okay, looks good so far. I am feeling good, so it's time to launch Cyberpunk 2077, which was causing my card to become frighteningly toasty at 86°C and up. I didn't change the settings and oh my, it barely hits 60°C in game now!
So, the conclusion of my saga with the overheating RTX 2070 is this: repaste and repad an aging card and you should see massive temperature and performance increases. However, only do this if you're comfortable with potentially voiding a warranty and have previously applied thermal paste to a chip (like a CPU). Graphics cards are expensive to replace...